Comparing exp() performance on Julia versus numpy

General Usage Performance
15

Basically the reason is that Julia can do this if it is able to inline the code, but generically turning scalar code into vector code through the compiler only is a very complicated optimization.